MultiSense sensors are stereo cameras, so they measure range by measuring the offset of features in the left and right camera. The offset is smaller as features are further away, so error in the measurement increases with distance. At large distances the offset cannot be measured, so an infinite range is reported by the camera.
The theoretical accuracy of a stereo camera depends on the distance between the two cameras (baseline), and the focal length of the lenses. The larger the baseline, the more offset there is in the left and right images. Longer focal lengths have the effect of "zooming in", which reduces the FOV, but gives better detail at further distances.
The actual accuracy of the camera also depends on things such as the quality of the calibration, the quality of the images, and the surface being viewed. Errors in each step of the rectification and stereo matching process accumulate to create small errors in the output images.
Before shipping each camera, we check its accuracy by placing a planar target at various known distances and plot the distance measured by every pixel at each distance. We then make sure that the median range error is less than half of a pixel at every range. Keep in mind, there can be noise in the measurements that have more than 1/2 pixel of error, and we only measure and test ranges up to 10 meters before shipping.
Comments